Methods In The Design And Implementation Of The Randomized Evaluation .

Curley et al. Trials(2018) 19:687Page 3 of 10Table 1 Randomized Evaluation of Sedation Titration forRespiratory Failure (RESTORE) eligibility criteria Enhance treatment fidelity by linking the arousalInclusion criteria: Rename the “plateau phase” the “titration phase” to 2 weeks of age, 42 weeks post-menstrual age, and 18 yearsof age Supported on mechanical ventilation for acute lung disease. Lungdisease includes both airways and parenchymal diseaseExclusion criteria: Cyanotic heart disease with unrepaired or palliated right to leftintracardiac shunt History of single ventricle at any stage of repair Congenital diaphragmatic hernia or congenital/acquired diaphragmparalysis Primary pulmonary hypertension Critical airway (e.g., post laryngotracheal reconstruction) oranatomical obstruction of the lower airway (e.g., mediastinal mass) Ventilator dependent (including non-invasive) on PICU admission(chronic assisted ventilation) Neuromuscular respiratory failure Spinal cord injury above the lumbar region Pain managed by patient-controlled analgesia (PCA) or epiduralcatheter Patient transferred from an outside ICU where sedatives had alreadybeen administered for more than 24 h Family/medical team have decided not to provide full support(patient treatment considered futile) Enrolled in any other critical care interventional clinical trialconcurrently or within the last 30 days Known allergy to any of the study medications Pregnancy[30–34]. Arousal assessments were performed in 13 patients and lasted 290 min (median; IQR 75–2160 min).One patient did not awaken during an assessment,which prompted a neurological evaluation that identifiedan intracranial hemorrhage. There were also trends toward decreased PICU and hospital lengths of stay, lessopioid exposure and fewer days of opioid exposure during the intervention phase.Based on these pilot data and clinician debriefingafter the pilot, we made the following changes toRESTORE-beta: Improve care provider buy-in and facilitateimplementation by assuring medical and nursingPICU leadership support using a signed contractshowing support Improve protocol compliance by emphasizing theRESTORE goal to ensure an awake yet comfortablepatient using the lowest effective dose of a limitednumber of analgesics and sedatives for the shortestperiod of timeassessment to the patient’s sedation levelreinforce that the goal of that phase of the protocol isto titrate, not maintain comfort medication infusions Avoid confusion in drug titration by simplifying thealgorithm by matching the frequency of drugtitration (every 8 h) in the titration and weaning toextubation phases Assure compliance with extubation assessment bymodifying the extubation readiness test when thepatient is spontaneously breathing with anoxygenation index of 6. (Use a SpO2 – EstimatedPaO2 conversion table to estimate the PaO2 inpatients without an arterial line.) Make dose titration consistent by changing the rateof opioid weaning from 20% every 12 h to 10% every8 h in the weaning phaseRESTORE – Phase III clinical trialDuring study design, we rejected the idea of performing abefore/after trial as we did in the pilot because changes inusual PICU care could occur over time and impact studyoutcomes. We also considered randomization by patient,team and PICU. We rejected patient-level randomizationbecause the intervention required a PICU-wide practicechange in how clinical teams work together. Patientrandomization could lead to group contamination overtime and be disruptive to unit operations since bedsideteams could be responsible for caring for patients in eachgroup at the same time. We rejected team randomizationbecause multiple teams rotate in a PICU over the courseof a day. We chose cluster randomization by PICU to limitcontamination between groups. In addition, a multicenterclinical trial allows potential bias to be distributed acrossdiverse practice settings, allows the comprehensive assessment of patient risk, allows the determination of the levelof protocol compliance necessary to assure desired patientoutcomes and increases the generalizability of studyresults. We therefore developed the multicenter, clusterrandomized clinical trial testing the RESTORE protocol onthe duration of mechanical ventilation in pediatricpatients supported on mechanical ventilation for acuterespiratory failure (U01 HL086622 and HL086649; Curleyand Wypij).We selected a parallel cluster randomized clinical trialdesign rather than a stepped-wedge design because (1)there were no data to support the RESTORE intervention and (2) PICUs were hesitant to implement an untested intervention. The hybrid stepped-wedge clusterrandomized trial (intervention delivered in half thePICUs randomized to the intervention in a steppedwedge fashion with the remainder functioning as controls) could have been selected as an alternative to the

Curley et al. Trials(2018) 19:687parallel cluster study though we predicted that thePICUs were relatively homogeneous with a small intracluster correlation coefficient (ICC) so that our parallelstudy would tend to deliver better statistical performance than a stepped-wedge trial. Logistically, we werealso prepared to start all intervention sites at approximately the same time.Initially we designed the study using the PICU as theunit of randomization where all patients in the randomized PICU would receive the intervention and providedeidentified information. However, during this time, theDepartment of Health and Human Services’ Office ofHuman Research Protection (OHRP) produced opinionsaround risks to subjects enrolled in clinical trials [35].Based on the newly published recommendations, the RESTORE Data and Safety Monitoring Board (DSMB)asked that we contact OHRP to opinion on our plannedstrategy and need for informed consent. OHRP thenopinioned that RESTORE was a research interventionthat posed more than minimal risk and that all parents/legal guardians should provide informed consent. Therefore, RESTORE was implemented as a parallel clusterrandomized controlled trial where PICUs assigned to theintervention group implemented the RESTORE protocolas a research intervention only on consented subjectswhile PICUs randomized to the control group continuedtheir usual sedation practices and data were collectedonly on consented subjects.The RESTORE intervention was an organizationalchange directed at all PICU clinicians and included:Page 4 of 10each PICU to identify potential subjects. Daily data collection occurred in all enrolled subjects from endotrachealintubation until the end of their scheduled sedation therapy, hospital discharge, or day 28 (whichever occurredfirst). Half of all enrolled subjects continued to be followedat approximately 6 months post PICU discharge for comparison of long-term outcomes. The primary outcome ofthis Phase III clinical trial was the duration of mechanicalventilation. Secondary outcomes are shown in Table 2.Collaborating centersClinical sites were recruited from the Pediatric AcuteLung Injury and Sepsis Investigators (PALISI) Network.Participating PICUs mock screened their unit for 1month to evaluate their available patient population andcompleted an organizational assessment describing theirparticipating unit structure, work processes, change processes and sedation practices. Organizations with morethan one PICU selected one participating PICU thathoused their general medical patients. All participatingPICUs (1) verified that they did not have a sedationprotocol in place, (2) showed evidence of critical carenursing and physician leadership support, (3) agreed toimplement the same pain, sedation and withdrawalassessment instruments and (4) could enroll a minimumof three subjects per month. Because of the cluster randomized design, the participants needed to agree to theTable 2 Randomized Evaluation of Sedation Titration for RespiratoryFailure (RESTORE) secondary outcome variables Time to recovery of acute respiratory failure (from endotrachealintubation to first meeting criteria to be tested for extubation readiness)1. Team education on the use of the sedation protocol inpediatric patients supported on mechanical ventilation2. Team identification of the patient’s trajectory ofillness and daily prescription of a sedation goal3. A nurse-implemented, goal-directed, comfortalgorithm that guides moment-to-momenttitration of opioids and benzodiazepines4. Team feedback on sedation management performanceThe unit of randomization was the PICU, the unitof inference was the patient and we controlled forcenter effects.The study was reviewed and approved by the DSMBand all local Institutional Review Boards, including thoseof the clinical and data coordinating centers. Legalguardians of all study participants provided writteninformed consent using one of three separate consentdocuments (baseline phase, usual care, or intervention)that described the cluster randomized design and therisk associated with their level of participation (datacollection with/without intervention). Inclusion andexclusion criteria were unchanged from the pilot study(Table 1). Screening was performed at least once daily in Duration of weaning from mechanical ventilation (from first meetingcriteria to be tested for extubation readiness to first successfulextubation – defined as extubation for more than 24 h) Occurrence of adverse events: inadequate pain management,inadequate sedation management, clinically significant iatrogenicwithdrawal symptoms, unplanned extubation, airway irritation frommovement of the endotracheal tube within the airway, extubationfailure/reintubation within 24 h of extubation, dislodgement of vascularaccess or drainage tubes, ventilator-associated pneumonia (VAP)a,catheter-associated blood stream infection (CA-BSI)a, and stage 2 pressure ulcers. Report of a new critical airway will be assessed throughhospital discharge or day 90 (whichever occurs first) Detection of life-threatening neurological events Occurrence of iatrogenic withdrawal symptoms PICU and hospital LOS Hospital costs Protocol implementation costs Cost-effectiveness In-hospital mortality Post-discharge quality of life and emotional healthaThe National Nosocomial Infections Surveillance System (NNIS) definitions willbe used to define VAP and CA-BSI. All cases of VAP and CA-BSI will beadjudicated by a process outlined by Cook et al. [47]. LOS length of stay, PICUpediatric intensive care unit

Curley et al. Trials(2018) 19:687research design where PICUs would be randomized tothe RESTORE intervention as a research protocol whilethe remaining PICUs continued usual care and wouldcontinue to maintain their current practices in thecontrol PICUs. During the pre-randomization period,all PICUs provided baseline data on their usual sedationpractices and provided approximately 3 months of baseline data so that site comparability at trial entry could beevaluated. A natural grouping of PICUs emerged based onthe number of eligible patients: small, medium and large.In 2009, the study began with 22 sites randomized(12 intervention and 10 control). Because of the needto obtain consent in control PICUs where only datacollection was occurring versus the intervention unitswhere a change in practice was being performed, weanticipated lower consent rates in the interventionPICUs. Based on our previous experience, we anticipated a consent rate of 85% in control PICUs and aconsent rate of 60% in intervention PICUs [36, 37]. Because of these estimates, we randomized more sites tothe intervention arm. Study enrollment was subsequently reviewed in June 2010. Enrollment rates fromthese 22 sites were slightly lower than expected. To increase enrollment rates, we added nine sites in 2010,following the same site selection procedures andrandomization scheme as for the original sites. Specifically, we recruited them from the PALISI Network, performed an organizational assessment, classified them bysize and randomized them by size, five to the intervention arm and four to the control arm.Study proceduresStudy procedures were focused to enhance treatmentfidelity as follows. Each participating center identified aphysician-nurse-pharmacist co-investigator team responsible for discipline-specific education, complianceassessment and retraining. After randomization, co-investigators from the intervention sites attended astart-up meeting and completed a competency-basedtraining program and certification process prior to enrolling intervention patients. This process included review of the web-based Manual of Operations and studyvideos.After the start-up meeting, intervention PICUs conducted team training for 1 month and then implementedthe RESTORE sedation protocol on consented patients.Team training included all clinicians (physicians, nurses,clinical pharmacists, respiratory therapists) involved in themanagement of intubated and mechanically ventilated patients. Training materials included discipline-specific slidepackages, informal case discussions, a video of the nurseimplemented, goal-directed, comfort algorithm andarousal assessment, pocket reminder cards and bedsidebooklets. Prior to the intervention phase, all physicians,Page 5 of 10unit-based clinical pharmacists and nursing staff were required to demonstrate their understanding of the intervention by completion of a discipline-specific, scenariobased self-assessment evaluation requiring a score of 80%. If a score was 80%, retraining was required, andthe assessment repeated until the score was 80%. Inaddition to the core physician-nurse-pharmacist team,additional multidisciplinary “Champions” served as unitbased resources for the RESTORE protocol. Championsprovided clinicians access to a RESTORE protocol expertat all times.During routine daily multidisciplinary patient care rounds,the protocol directs clinicians to identify the patient’s trajectory of illness and prescribe a daily sedation goal. The bedside nurse then used the algorithm with complementarystandardized order set to titrate the comfort medications to the prescribed sedation target. The order setserved three purposes: (1) reinforcement of training,(2) decreased delay in implementing a change in sedation and (3) enhanced protocol compliance.The decision to require consent for each interventionpatient impacted the trial by providing staff utilizing RESTORE with less experience using the protocol. Becauseof this we required a minimal enrollment of three subjectsper month and a rigorous training and quality improvement plan. To assure protocol compliance, reinforce education and assess safety during the intervention period,one site co-investigator rounded separately on each studypatient each day. During these investigator rounds, theco-investigator or nurse champion offered staff supportand retraining as necessary and completed a “WalkRounds Report” that collected data around these topicsand issues. Any deficits identified during walk rounds wasaddressed with the care team to assure compliance withthe protocol. The reports were summarized weekly to provide team feedback on RESTORE performance.Post-discharge quality of life and emotional healthFollow-up was conducted on a stratified random sampleof approximately half of the consented subjects. To ensure that the sample was representative of all subjects inthe trial, the sample was stratified by study site and age.Consenting families were sent a reminder letter and copies of the follow-up assessment instruments by mail andcalled 6 months ( 1 month) after PICU discharge.We monitored our follow-up rates monthly and increased our sampling ratio when necessary to meet expected milestones. Follow-up rates did not differ bytreatment arm. Treatment group differences in patientswith follow-up were similar to those in the main trial,with patients in the intervention arm being younger andhaving lower risk of mortality, less frequent history ofasthma and a different distribution of primary diagnosiscategory [38].

Curley et al. Trials(2018) 19:687Statistical considerationsThe PICUs were block randomized by size (small,medium and large) to assure a balanced allocation between the control and intervention groups.With respect to the primary outcome, duration ofmechanical ventilation, subjects were assigned 28 daysif they were still intubated after 28 days or if they weretransferred or died prior to day 28 without remainingextubated for 24 h prior to transfer/death. Thismortality-adjusted duration of mechanical ventilation isa continuous variable that is effectively equivalent toventilator-free days [39]. We anticipated that the mortality rate in the first 28 days would be low and similarbetween the control and intervention groups. If thatwas not the case in the study, we would have performeda secondary analysis excluding these deaths from theanalysis, effectively comparing the duration of mechanical ventilation among survivors. If the mortality ratewas higher than anticipated and/or is unbalanced between control and intervention groups, we planned toconduct a competing risks analysis, treating extubationand death as two competing events.The primary analysis compared the duration of mechanical ventilation in intervention and control subjectsusing Kaplan-Meier survival curves and proportionalhazards regression adjusting for age group, PRISM IIIscore and POPC at enrollment. We considered PICU asa cluster variable in the survival analyses using Lin andWei’s sandwich variance estimator [40].Study monitoringAn independent DSMB monitored the clinical trial foradverse events, adherence to study protocol and potential early stopping. To determine if early stopping waswarranted, group sequential monitoring was used to assess for efficacy and the method of stochastic curtailment was used to assess for futility. The sample size wasadjusted using an O’Brien-Fleming stopping rule to accommodate three formal interim analyses after approximately 400, 1200 and 1800 subjects [41].Sample size and power considerationsWe hypothesized that patients managed with a pediatricsedation management protocol would experience ashorter duration of mechanical ventilation compared withthose not managed with the sedation protocol. The studyteam determined that a 20% reduction in the duration ofmechanical ventilation was clinically important. This clinically important 20% reduction, or hazard ratio of 1.25,was plausible based on the results of the pilot study.Assuming independent observations and proportionalhazards between treatment groups, 892 events (successful extubations) were required for a two-sided, 0.05-levellog-rank test to achieve 90% power to detect a 20%Page 6 of 10reduction in ventilation duration with the interventionassuming three interim analyses (East, Version 5.3, CytelStatistical Software, Cambridge, MA, USA). As we expected no more than 15% of the subjects to be censoredat 28 days given our experience with the prone positioning study [37] and the pilot sedation management study,we required 1050 subjects to achieve this power. However, because observations from subjects at the same sitemay be correlated, we inflated this sample size toaccount for the ICC in our cluster randomized trial design [42]. The ICC for the Martingale residuals of extubation times was estimated to be 0.00 in the seven-siteprone-positioning study [37] and 0.01 in the two-sitepilot sedation management study. Conservatively basedon the first 22 RESTORE sites, using ICC 0.01 leads to1990 total subjects needed. Increasing the number ofsites or number of subjects will increase the power. Wechose 2448 enrolled subjects as our target sample size,guaranteeing 90% power to detect a 20% reduction inlength of extubation controlling for censoring, three formal interim analyses for early stopping and modestwithin-site correlations.Quality controlPediatric critical care nurses with research experienceuniquely served as study monitors and conducted sitevisits, beginning at least 6 months after the start of theintervention phase of the trial. The design of the trial required that intervention sites implement the intervention on consented subjects and that control sites notimplement the protocol. We monitored the interventionin the intervention sites and

4Department of Pediatrics, Medical College of Wisconsin, Milwaukee, WI, USA 5Critical Care Section, Children's Hospital of Wisconsin, 9000 W. Wisconsin Avenue, MS #681, Milwaukee, WI 53226, USA Full list of author information is available at the end of the article